Marius ComperDashboard

Method & sources

← Back to the dashboard

Method & sources (D1–D6)

D1 · World Bank WDI (CC BY-4.0, attributed)

10 indicators: SP.POP.TOTL / NY.GDP.MKTP.CD / NY.GDP.PCAP.CD / NY.GDP.MKTP.KD.ZG / FP.CPI.TOTL.ZG / SP.DYN.LE00.IN / AG.LND.TOTL.K2 / IT.NET.USER.ZS / SP.URB.TOTL.IN.ZS / IT.CEL.SETS.P2 (ROU;CHN). All re-pulled at build: 9 complete 2024 values, area null for both countries in 2024 (verified), hence the 2023 value with explicit vintage. API template: api.worldbank.org/v2/country/ROU;CHN/indicator/{CODE}?date=2024&format=json. Data source: World Bank, World Development Indicators; license CC BY-4.0.

D2 · ECB rates (daily XML)

eurofxref-daily.xml, build-time vintage 2026-09-17: EUR/USD 1.1481, EUR/RON 5.2624, EUR/CNY 7.7009. Unit rate[X] = X per EUR; conversion = amount × rate[to] ÷ rate[from] → 1 RON = 1.4634 CNY, 1 CNY ≈ 0.68335 RON (tolerances on unrounded values). The fixing is not a sell rate. If the XML lags over 5 business days the converter freezes (this build: active).

D3 · Bilateral trade (BACI via OEC tesseract)

Cube trade_i_baci_a_92, Year cut + Year drilldown (rows carry Year), row-level check: each target row with Year = requested year and the aschn/eurou ID pair in the requested direction, else stop. 2024: China→Romania $8.78B / Romania→China $2.44B (29,875 rows, sha256 f296d0f4e980); 2023: China→Romania $7.46B / Romania→China $1.87B. The pitfall: a lowercase year= parameter is ignored, and an uncut query returns multi-year sums (about $90B/$25B, void, anti-fixture F4). BACI is reconciled academic data with slightly different methodology from customs.

D4 · Clocks (derived from IANA rules)

Badge gap(t) = off(Asia/Shanghai,t) − off(Europe/Bucharest,t). Rules (rechecked at build against the IANA europe/asia sources): the EU springs forward on March's last Sunday at 01:00Z and falls back on October's last Sunday at 01:00Z (2026: 3-29/10-25); Shanghai's PRC rules end in 1991, hence fixed UTC+8. The timeapi.io reading (2026-09-17T21:26Z, 5h) is an observation anchor only, not the logic source. Note: macOS system zoneinfo wrongly gave 04:00Z — the IANA sources rule.

D5 · Distance (self-computed haversine)

OTP/LROP 44.571792N 26.103285E → PEK/ZBAA 40.077349N 116.596702E (OurAirports database, retrieved 2026-09-17; Romania's AIP couldn't be obtained directly — honest ceiling: non-AIP values), R = 6371.0088 km → 7,054.2 km (0.5 km tolerance, recomputed at build).

D6 · The nonstop flight (verified, confirmed)

Since 4 September 2026, Air China Beijing–Bucharest CA913/CA914, 3× weekly (Mon/Wed/Fri), A330-200; timetable OTP→PEK about 9 hours 40 minutes, PEK→OTP about 10 hours 50 minutes (per September clocks). Outbound the same number continues to Zagreb, but both OTP–PEK legs are nonstop. Sources: the EX-YU Aviation timetable (body verified). Timetables change; rechecked per lifecycle.

Methodology note

Taiwan's trade.gov.tw Romania sheet (06.2026) gives GDP of about $356B versus the Bank's $382.6B: different methodologies, the Bank wins, no mixing. georank/CEIC are SERP-competitor records only, never evidence.

Lifecycle & circuit-breakers

Bank cards yearly (vintage over 18 months → “stale data” banner, over 24 → hidden); converter at each daily build (XML over 5 business days stale → frozen); clocks only on IANA rule changes (rechecked in transition weeks); OEC yearly (cut + row-level + magnitude checks every time); version-pinned snapshots (this one: 34bd882ab4ec), corrections with rollback + note; source links quarterly.